ORCID

https://orcid.org/0000-0002-1026-6940

Date of Award

Winter 12-9-2023

Author's School

Graduate School of Arts and Sciences

Author's Department

Psychology

Degree Name

Master of Arts (AM/MA)

Degree Type

Thesis

Abstract

People spontaneously segment continuous streams of experiences into distinct episodes. Prediction errors are theorized to drive segmentation. However, existing studies exploring the relationship between prediction error and event segmentation lack a continuous measure of prediction error during naturalistic perception and often fail to distinguish between prediction error and prediction uncertainty. Do moment-by-moment fluctuation of prediction errors, not prediction uncertainty during naturalistic event perception correlate positively with segmentation probabilities? To tackle this question, we harnessed the predictive nature of eye movements and introduced a predictive looking model. In this model, individuals’ prior gaze patterns act as predictors for critical features in the current frame. Testing the model using group gaze density maps from participants engaged in passive movie viewing—with actors’ hand locations serving as an approximation for the prediction target—we uncovered that past gaze patterns, up to 9 seconds prior, predict the current locations of actors’ hands in the movie. Furthermore, a significant and positive correlation emerged between predictive looking errors and prediction errors generated by a computational model. This suggests a congruence in capturing true prediction error signals in the brain. Crucially, aligning with theories proposing an association between increased prediction errors and event segmentation, predictive looking errors positively correlated with event segmentation probabilities.

Language

English (en)

Chair and Committee

Jeffrey Zacks

Committee Members

Zachariah Reagh, Richard Abrams

Download

Share

COinS